Add lock wrapper for Channels.

[TommyXR]

lock(f, lck) is defined for locks, but not for Channels. I think it's simply a mistake, thus I added lock(f, ::Channel) for API uniformity.

[TommyXR]

Some people on the slack have argued that we should instead remove the ::ReentrantLock type restriction in the other file.

I don't mind changing the commit to reflect this, but I would like to know what people think is the better option.

[DilumAluthge]

I don't feel strongly either way.

[JeffBezanson]

Duplicating the code is not good. There are two other methods that use this approach:

lock(f, c::GenericCondition) = lock(f, c.lock)

That's better since there's no duplication, but in fact we could just remove the ::AbstractLock restriction and remove the existing extra methods.

More weirdness:

• lock(::WeakKeyDict) is not defined but should be (also unlock)
• unlock(f, c::GenericCondition) = unlock(f, c.lock) is defined, but there is no similar unlock method for it to call (?)

[TommyXR]

Duplicating the code is not good. There are two other methods that use this approach:

lock(f, c::GenericCondition) = lock(f, c.lock)

That's better since there's no duplication, but in fact we could just remove the ::AbstractLock restriction and remove the existing extra methods.

More weirdness:

* `lock(::WeakKeyDict)` is not defined but should be (also `unlock`)

* `unlock(f, c::GenericCondition) = unlock(f, c.lock)` is defined, but there is no similar unlock method for it to call (?)

I actually missed the fact that Channel.cond_take was actually a condition and that the method existed for GenericCondition. Made the change. I also removed unlock(f, ::GenericCondition) because after looking for a bit it's definitely a mistake and I added the missing function for WeakKeyDict.

[JeffBezanson]

It looks like we are also missing lock(f, ::IO). That makes me think it's better just to remove all but one of the 2-argument lock methods, and define lock(f, l::Any).